I am a postdoc researcher at the University of Aveiro. My expertise focuses on developing efficient Real-Time Communication Systems (RTCS), including Software-Defined Networking (SDN), Time-Sensitive Networking (TSN), Wi-Fi 6 & 7 enhancements, and Quality of Service (QoS) optimization for industrial applications.
This work presents an IT/OT converged testbed integrating Wi-Fi 7 and TSN to achieve ultra-low latency, high reliability, and enhanced network efficiency for real-time industrial applications. To achieve these objectives, wireless connections have been upgraded to Wi-Fi 7 technology to evaluate its feasibility for industrial use cases. For instance, AGVs have been redesigned to enhance mobility and enable seamless handovers in a Wi-Fi 7-enabled network, ensuring reliable connectivity in high-interference environments. In the second phase, legacy non-real-time industrial networks have been upgraded to TSN-enabled infrastructures, enabling deterministic communication with high reliability, enhanced QoS, and efficient data exchange.
This work extends the previous work by refining the formal system model of RT-MQTT, complementing the previous analytic approach based on the Holistic Approach (HA) with the more efficient Trajectory Approach (TA) to improve the tightness of the results. The work presents a comprehensive study of the application of response time analysis to RT-MQTT on multi-hop SDN/OpenFlow switched networks, with a focus on the Trajectory Approach for non-preemptive fixed-priority scheduling of sporadic messages, comparing its results with those obtained using the Holistic Approach. The analytic results are verified through extensive empirical testing within the Mininet emulator framework, being determined that the RT-MQTT framework is predictable and analyzable, that both TA and HA provide safe upper-bounds to network-level latency, and that the TA presents a lower level of pessimism, compared to the HA, at expenses of higher complexity.
This work developed a new architecture referred to as MRT-MQTT (Multicast Real-Time MQTT) that leverages multicast-based connectivity across multiple edge networks that adhere to the RT-MQTT architecture. This paradigm brings data processing and analysis closer to the network edge, aligning with the focus of several research and development efforts such as the MIRAI project. In MRT-MQTT, applications running on edge networks can explicitly state their real-time requirements. These requirements are then utilized by the SDN/OpenFlow resource manager to establish real-time channels within and between the edge networks. This architecture relies on traffic segregation to prioritize time-sensitive data flows over non-time-sensitive ones, thereby enhancing the timeliness of critical data. Multicast routing is applied between the edge networks to reduce the overall network load, consequently reducing latency and jitter values in real-time data communications. A reduction in transmission delay is achieved, amounting to 29% and 23% for QoS=0 and QoS=1, respectively, in comparison to DM-MQTT (Direct Multicast-MQTT), and 55% and 43% compared to standard MQTT. In the latter scenario, MRT-MQTT also led to a substantial decrease in network usage, registering reductions of 58.0% and 45.0% for QoS=0 and QoS=1, respectively.
This work developed an SDN-based resource management framework to enhance flexibility in network management. The framework enables the handling of device and protocol heterogeneity while satisfying real-time requirements, specifically allowing deterministic real-time data transmissions. These requirements and constraints are increasingly prevalent in industrial environments. This framework leverages MQTT, incorporating a set of real-time extensions known as RT-MQTT (Real-Time MQTT) for enhanced functionality. RT-MQTT empowers applications to specify real-time MQTT requirements, which are then translated into SDN/OpenFlow-enforced network reservations to establish real-time channels. By employing various combinations of network topologies and load levels and comparing them to scenarios without the proposed real-time mechanisms, a reduction of both average and worst-case latencies for time-sensitive traffic to approximately half is observed. In contrast, for normal traffic, there was an increase of approximately 10%.
Dr. Ehsan Shahri
IT - Instituto de Telecomunicações
University of Aveiro
Aveiro, Portugal
Email: ehsan.shahri@ua.pt